Sony AI Unveils Ace, First Robot to Defeat Elite Human Table Tennis Players

Sony AI announced in a press release the debut of Ace, an autonomous robot that can compete and win against elite and professional human table tennis players. The research, published in Nature, marks the first documented case of a robot achieving expert-level performance in a competitive physical sport.

Ace integrates Sony's sensor technology, reinforcement learning, and precision robotic hardware. It uses nine active pixel sensor cameras and three event-based vision systems to track the ball's 3D position, spin, and velocity in real time. The control system relies on model-free reinforcement learning, allowing it to adapt quickly without pre-programmed models.

According to the results reported in the Nature study, Ace played matches under official International Table Tennis Federation regulations, winning multiple games against elite players and achieving a 75 percent return rate on spins up to 450 radians per second. The robot also demonstrated low-latency perception and control, enabling it to respond effectively to unpredictable shots.

Subsequent matches in December 2025 and March 2026 showed continued improvement, with Ace defeating additional professional players and increasing both shot speed and rally pace. Sony AI stated that the project demonstrates how physical AI can perceive, plan, and act at human reaction speeds in complex, dynamic environments.